SQL Server Adventureworks SalesOrderDetail表主键

Question

I would like to ask about the actual reason using the field SalesOrderId as part of the primary key. I think that this should be only a foreign key to the table SalesOrderHeader .

Thanks in advance

Answer 1

Ok a little about indexes There are mutiple types. There is a clustered index which determines the physical order the data has on the disk, without this the table is technically known as a heap. Access of an large volume of data is inefficient. The data is random. Any non clustered index (and statistics about data distribution) will refer to the custered index to access data. When an index is built the query optimiser will use indexed columns from the left in order to find data. So when there is an index col1, col2, col3 if I am joining, searching or whatever on col1, the index will be used. If Im filtering on col1,col2 the same index will be used. There is little or no penalty for this. If my query references col1, col2, col3 ONLY (including where and returned columns etc) the the information to be returned can be acertained by ONLY reading the index, the table its self does not need to be accessed. This is known as a covering index. So this is very quick. In general, tables should be narrow, indexes wide. To find out more I recommend http://www.insidesqlserver.com/thebooks.html Kalen Delany: she knows the SQL dev team personally and writes very clearly. The basic principles are still true from SQL7 through to SQL2012 - although sparce indexes, spatial data types, BI and a lot of other goodies have been added.